[NTLUG:Discuss] Video Card Recommendation? -- Long Redux
Bryan J. Smith
b.j.smith at ieee.org
Sat May 8 09:35:12 CDT 2004
Steve Baker wrote:
> It's not quite fair to call their OpenGL extensions "non-standard".
> ...
Okay, you're correct in that regard. But it does seem that it's been
far too long between OpenGL 1.3 and the proposed OpenGL 2.0 work.
> Hence, there is no excuse for anyone to fail to support a new core
> feature in their latest hardware - they can see it coming at least a
> year or so ahead of time. This is a good way to proceed. All OpenGL
> implementations are required to support ALL of the core API. This is
> a welcome contrast to Direct3D where you have to test all manner of
> 'capability bits' to determine whether even important core facilities
> are present.
Exactomundo. And the fact that various capabilities are always
available in software in OpenGL, whereas DirectX does not.
> Yeah - but they don't even provide OpenGL drivers for Linux. If you
> want to use 3DLabs hardware to do 3D graphics under Linux, you have to
> pay money to a 3rd party company for a customised X server and their
> driver (which is crap and doesn't support the latest 3D labs cards
> anyway).
3DLabs seems to be releasing the source code for Linux drivers a
_number_ of their newer Wildcat cards. Yes, they are newer to Linux,
but they have been moving forward in the past 2 years.
> Sure it is! ATI have been working hand-in-hand with 3D labs on the
> 2.0 specification. ATI have had working drivers for about six
> months. nVidia have been dragging behind because the OpenGL 2.0
> shader language (GLSL) is a competitor to their own 'Cg' shader
> language...but the latest Windows drivers for nVidia/OpenGL do
> actually support GLSL - although it has to be turned on with a secret
> registry hack because it's still 'in beta test'.
Ahhh, thanx for that insider knowledge. I appreciate it. My info is
about 2 years old.
As far as Cg, I thought nVidia was freely sharing it. Is this not true?
> I expect to see non-beta OpenGL 2.0 drivers from both ATI and nVidia
> in time for SigGraph this year.
Very cool indeed!
> ATI have shared driver sources under heavy NDA's with a number of
> companies (Evans & Sutherland for example). Meanwhile, nVidia stopped
> providing even their own board manufacturers with source code.
Ouch! Did not know that about nVidia. Again, my info is dated about 2
years now.
> Only for the 2D interfaces - there is no 3D info for anything more
> recent than TNT.
Right. But at least the full 2D specs are there, or at least that's
what I've seen.
> Not exactly. It was HEAVILY obfuscated sources - quite utterly
> unreadable. No better then disassembled machine code.
Right, but then the community started reworking it and putting back in
the formal identifiers and that's when the 3rd parties started bitching
and moaning about litigation. It's quite sad.
I just dislike all the "demonization" of nVidia by the community.
Without their continued release of Linux drivers, Linux would be a piss
poor OpenGL platform in general.
> The main problem with ATI's OpenGL drivers is that they are just buggy
> as hell. Really awful. We get repeated crashes, freezes, incorrect
> adherence to the OpenGL spec, memory leaks...you name it.
And that's the problem with creating a "clean room" implementation.
You have to give ATI some credit for trying, but it seems that even
they agree with nVidia. It's best to just use what you got already,
even if that means closed source.
Thanx for the update on a few items. Glad to see OpenGL 2.0 work _is_
progressing.
--
Bryan J. Smith, E.I. -- Engineer, Technologist, School Teacher
b.j.smith at ieee.org
More information about the Discuss
mailing list